As a meteorologist, I am used to dealing with uncertainty. The atmosphere is complex and our models always paint an incomplete, but still useful picture of the weather. We strive to be as confident in our conclusions as possible, but are careful to avoid showing over-confidence by drawing conclusions that are simply not justified by the data itself.

I was therefore troubled by Matt Rogers’ recent Capital Weather Gang post, “Global warming of the Earth has decelerated.” A more accurate title might have been “A partial look at recent temperature data trends without quantifying significance or uncertainty.”

Ok. That title isn’t the catchiest, I admit, but it would have been more accurate.

While others have already taken a discerning look at some of the analysis’ technical flaws (see here and here), I want to focus on our ability to tease out conclusions in data as inherently noisy as global surface temperatures.

As you may have noticed, weather can be a bit chaotic. That can extend to the entire planet, where human-induced climate change from greenhouse gas emissions is operating right alongside natural variability. This is why climate scientists look at longer time frames (at least ~ 30 years) to determine trends. Effects can occur at smaller timescales, however, and that motivates plenty of research as well. And so we find ourselves here—discussing what some have called a “hiatus,” a “speed bump” or a “faux pause.” For his part, Rogers labels it a “deceleration.”

Rather than look at how global surface temperatures have changed over time (they’re increasing), Rogers computed how quickly global temperature has been rising then calculated a trend line that showed a slight downward trend in the rate of warming during a relatively short time frame. From this, he concluded that, “… if current trends continue for just a few more years, then the mean change for the 2000s will shift to negative; in other words, the warming would really stop.”

That conclusion is incredibly problematic. Because we’re talking about acceleration here (a change in the rate), it’s akin to claiming that a driver who slows down as he or she gets stuck in traffic will never get to his/her destination. To put a finer point on it, nobody who drives from D.C. to New York City and gets stuck in Baltimore traffic just gives up and goes home.

Further, the evidence Rogers uses to make these statements are trend lines on a 13-year dataset. He did not share any test for statistical significance, nor an accounting of the inherent uncertainty prevalent in temperature time-series—where, importantly, weather variability can overwhelm climate signals, especially over such a short time-frame.

On climate, the blogger and statistician Tamino has several excellent posts illustrating this exact point.  In his post Uncertain T, he calculates global temperatures trends (i.e. the annual rate of change in temperature in degrees celsius) using different starting dates, ranging from 1975 to 2005. Although they use different data sets and metrics, this exercise largely mirrors the method used by Rogers. However, Tamino also included 95% confidence intervals on each trend line rate. These confidence intervals allow scientists to attach varying levels of certainty to their results. You can easily see that as the number of years in the dataset decrease, the 95% confidence intervals grow larger and larger. This makes sense. The less data you use for your analysis, the less confidence you will have in your results.

For additional description, see Uncertain T  (Tamino)

Tamino also included a dotted line indicating the calculated trend rate since 1975. At no point is the rate of warming using the 1975 start date not within the range of possibilities calculated based off different starting years. Even the 13-year trend from 2000 to 2013 – the range Rogers used — has confidence intervals that would allow for the true trend to range all the way from negative to positive, a result that suggests it’s not a very useful time frame to examine.

So what does this mean? In contrast to his assertion of “Verifying the pause,” Rogers has simply provided incomplete information, on which it is impossible to draw firm conclusions.

But what about Rogers’ claims that climate scientists themselves have acknowledged the pause? In his post’s comments, Rogers notes that Dr. Kevin Trenberth and Dr. Michael Mann are “fully admitting [sic] the pause and are conducting research to account for it.” This didn’t mesh with my readings of their research. So I asked them.

Dr. Mann responded thusly:

“It is [a] misrepresentation of my work to suggest that I have ‘admitted to a pause.’ I have, in fact, argued the opposite (that global warming continues unabated, even if natural internal variability may have offset some warming during the past decade).”

Dr. Mann also directed me to his recent Scientific American piece about what he calls this “Faux Pause.”

Similarly, Dr. Trenberth told me his work is “in marked contrast to the blog.” He was quick to point out that a straight line fit from 1970 to 2013 is still pretty good and while there has been a slowdown in the rise of global mean temperature, there has been no “pause” in the melt of Arctic sea ice, or global sea level rise, or global energy imbalance with the excess energy going deeper into the oceans.

Scientists can be many things, and curious is chief among them. But scientists also have a responsibility to communicate carefully. We must be disciplined in drawing conclusions that the data support, even when our curiosity demands more. And we need to be transparent about how confident we can be with the conclusions we do make.

Blog posts are great for exploring ideas, but they shouldn’t be used as an alternative venue in which to advance claims that aren’t strong enough to withstand scrutiny from other scientists.

As for climate change itself, I’m just a meteorologist. That’s why I go to scientific assessments from climate researchers like these, or these, or these, or these for their conclusions. They’re the experts. But if they want a five to ten-day forecast, they know who to call.

Tom Di Liberto is a meteorologist and science communicator who was named America’s Scientist Idol in 2013. The views expressed here are the author’s alone. Follow him on Twitter @TDiLiberto.